mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Mark Fasheh <mfasheh@suse.com>
To: linux-kernel@vger.kernel.org
Cc: ocfs2-devel@oss.oracle.com, Joel Becker <joel.becker@oracle.com>,
	Mark Fasheh <mfasheh@suse.com>
Subject: [PATCH 47/56] ocfs2: Use ocfs2_cp_xattr_bucket() in ocfs2_mv_xattr_bucket_cross_cluster().
Date: Mon, 22 Dec 2008 13:48:28 -0800	[thread overview]
Message-ID: <1229982517-3455-48-git-send-email-mfasheh@suse.com> (raw)
In-Reply-To: <1229982517-3455-1-git-send-email-mfasheh@suse.com>

From: Joel Becker <joel.becker@oracle.com>

The buffer copy loop of ocfs2_mv_xattr_bucket_cross_cluster() actually
looks a lot like ocfs2_cp_xattr_bucket().  Let's just use that instead.
We also use bucket operations to update the buckets at the start of each
extent.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
---
 fs/ocfs2/xattr.c |  169 +++++++++++++++++++++++++++++++++---------------------
 1 files changed, 104 insertions(+), 65 deletions(-)

diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
index 5efcf4e..5be9966 100644
--- a/fs/ocfs2/xattr.c
+++ b/fs/ocfs2/xattr.c
@@ -170,6 +170,11 @@ static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
 
 static int ocfs2_delete_xattr_index_block(struct inode *inode,
 					  struct buffer_head *xb_bh);
+static int ocfs2_cp_xattr_bucket(struct inode *inode,
+				 handle_t *handle,
+				 u64 s_blkno,
+				 u64 t_blkno,
+				 int t_is_new);
 
 static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb)
 {
@@ -3526,13 +3531,21 @@ out:
 }
 
 /*
- * Move half nums of the xattr bucket in the previous cluster to this new
- * cluster. We only touch the last cluster of the previous extend record.
+ * prev_blkno points to the start of an existing extent.  new_blkno
+ * points to a newly allocated extent.  Because we know each of our
+ * clusters contains more than bucket, we can easily split one cluster
+ * at a bucket boundary.  So we take the last cluster of the existing
+ * extent and split it down the middle.  We move the last half of the
+ * buckets in the last cluster of the existing extent over to the new
+ * extent.
+ *
+ * first_bh is the buffer at prev_blkno so we can update the existing
+ * extent's bucket count.  header_bh is the bucket were we were hoping
+ * to insert our xattr.  If the bucket move places the target in the new
+ * extent, we'll update first_bh and header_bh after modifying the old
+ * extent.
  *
- * first_bh is the first buffer_head of a series of bucket in the same
- * extent rec and header_bh is the header of one bucket in this cluster.
- * They will be updated if we move the data header_bh contains to the new
- * cluster. first_hash will be set as the 1st xe's name_hash of the new cluster.
+ * first_hash will be set as the 1st xe's name_hash in the new extent.
  */
 static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode,
 					       handle_t *handle,
@@ -3545,105 +3558,131 @@ static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode,
 {
 	int i, ret, credits;
 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
 	int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
 	int num_buckets = ocfs2_xattr_buckets_per_cluster(osb);
-	int blocksize = inode->i_sb->s_blocksize;
-	struct buffer_head *old_bh, *new_bh, *prev_bh, *new_first_bh = NULL;
-	struct ocfs2_xattr_header *new_xh;
+	int to_move = num_buckets / 2;
+	u64 last_cluster_blkno, src_blkno;
 	struct ocfs2_xattr_header *xh =
 			(struct ocfs2_xattr_header *)((*first_bh)->b_data);
+	struct ocfs2_xattr_bucket *old_first, *new_first;
 
 	BUG_ON(le16_to_cpu(xh->xh_num_buckets) < num_buckets);
 	BUG_ON(OCFS2_XATTR_BUCKET_SIZE == osb->s_clustersize);
 
-	prev_bh = *first_bh;
-	get_bh(prev_bh);
-	xh = (struct ocfs2_xattr_header *)prev_bh->b_data;
-
-	prev_blkno += (num_clusters - 1) * bpc + bpc / 2;
+	last_cluster_blkno = prev_blkno + ((num_clusters - 1) * bpc);
+	src_blkno = last_cluster_blkno + (to_move * blks_per_bucket);
 
 	mlog(0, "move half of xattrs in cluster %llu to %llu\n",
 	     (unsigned long long)prev_blkno, (unsigned long long)new_blkno);
 
+	/* The first bucket of the original extent */
+	old_first = ocfs2_xattr_bucket_new(inode);
+	/* The first bucket of the new extent */
+	new_first = ocfs2_xattr_bucket_new(inode);
+	if (!old_first || !new_first) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_read_xattr_bucket(old_first, prev_blkno);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+
 	/*
-	 * We need to update the 1st half of the new cluster and
-	 * 1 more for the update of the 1st bucket of the previous
-	 * extent record.
+	 * We need to update the 1st half of the new extent, and we
+	 * need to update the first bucket of the old extent.
 	 */
-	credits = bpc / 2 + 1 + handle->h_buffer_credits;
+	credits = ((to_move + 1) * blks_per_bucket) + handle->h_buffer_credits;
 	ret = ocfs2_extend_trans(handle, credits);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	ret = ocfs2_journal_access(handle, inode, prev_bh,
-				   OCFS2_JOURNAL_ACCESS_WRITE);
+	ret = ocfs2_xattr_bucket_journal_access(handle, old_first,
+						OCFS2_JOURNAL_ACCESS_WRITE);
 	if (ret) {
 		mlog_errno(ret);
 		goto out;
 	}
 
-	for (i = 0; i < bpc / 2; i++, prev_blkno++, new_blkno++) {
-		old_bh = new_bh = NULL;
-		new_bh = sb_getblk(inode->i_sb, new_blkno);
-		if (!new_bh) {
-			ret = -EIO;
+	for (i = 0; i < to_move; i++) {
+		ret = ocfs2_cp_xattr_bucket(inode, handle,
+					    src_blkno + (i * blks_per_bucket),
+					    new_blkno + (i * blks_per_bucket),
+					    1);
+		if (ret) {
 			mlog_errno(ret);
 			goto out;
 		}
+	}
 
-		ocfs2_set_new_buffer_uptodate(inode, new_bh);
+	/*
+	 * Get the new bucket ready before we dirty anything
+	 * (This actually shouldn't fail, because we already dirtied
+	 * it once in ocfs2_cp_xattr_bucket()).
+	 */
+	ret = ocfs2_read_xattr_bucket(new_first, new_blkno);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
+	ret = ocfs2_xattr_bucket_journal_access(handle, new_first,
+						OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret) {
+		mlog_errno(ret);
+		goto out;
+	}
 
-		ret = ocfs2_journal_access(handle, inode, new_bh,
-					   OCFS2_JOURNAL_ACCESS_CREATE);
-		if (ret < 0) {
-			mlog_errno(ret);
-			brelse(new_bh);
-			goto out;
-		}
+	/* Now update the headers */
+	le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -to_move);
+	ocfs2_xattr_bucket_journal_dirty(handle, old_first);
 
-		ret = ocfs2_read_block(inode, prev_blkno, &old_bh, NULL);
-		if (ret < 0) {
-			mlog_errno(ret);
-			brelse(new_bh);
-			goto out;
-		}
+	bucket_xh(new_first)->xh_num_buckets = cpu_to_le16(to_move);
+	ocfs2_xattr_bucket_journal_dirty(handle, new_first);
 
-		memcpy(new_bh->b_data, old_bh->b_data, blocksize);
+	if (first_hash)
+		*first_hash = le32_to_cpu(bucket_xh(new_first)->xh_entries[0].xe_name_hash);
 
-		if (i == 0) {
-			new_xh = (struct ocfs2_xattr_header *)new_bh->b_data;
-			new_xh->xh_num_buckets = cpu_to_le16(num_buckets / 2);
+	/*
+	 * If the target bucket is anywhere past src_blkno, we moved
+	 * it to the new extent.  We need to update first_bh and header_bh.
+	 */
+	if ((*header_bh)->b_blocknr >= src_blkno) {
+		/* We're done with old_first, so we can re-use it. */
+		ocfs2_xattr_bucket_relse(old_first);
 
-			if (first_hash)
-				*first_hash = le32_to_cpu(
-					new_xh->xh_entries[0].xe_name_hash);
-			new_first_bh = new_bh;
-			get_bh(new_first_bh);
-		}
+		/* Find the block for the new target bucket */
+		src_blkno = new_blkno +
+			((*header_bh)->b_blocknr - src_blkno);
 
-		ocfs2_journal_dirty(handle, new_bh);
+		/*
+		 * This shouldn't fail - the buffers are in the
+		 * journal from ocfs2_cp_xattr_bucket().
+		 */
+		ret = ocfs2_read_xattr_bucket(old_first, src_blkno);
+		if (ret) {
+			mlog_errno(ret);
+			goto out;
+		}
 
-		if (*header_bh == old_bh) {
-			brelse(*header_bh);
-			*header_bh = new_bh;
-			get_bh(*header_bh);
+		brelse(*first_bh);
+		*first_bh = new_first->bu_bhs[0];
+		get_bh(*first_bh);
 
-			brelse(*first_bh);
-			*first_bh = new_first_bh;
-			get_bh(*first_bh);
-		}
-		brelse(new_bh);
-		brelse(old_bh);
+		brelse(*header_bh);
+		*header_bh = old_first->bu_bhs[0];
+		get_bh(*header_bh);
 	}
 
-	le16_add_cpu(&xh->xh_num_buckets, -(num_buckets / 2));
-
-	ocfs2_journal_dirty(handle, prev_bh);
 out:
-	brelse(prev_bh);
-	brelse(new_first_bh);
+	ocfs2_xattr_bucket_free(new_first);
+	ocfs2_xattr_bucket_free(old_first);
+
 	return ret;
 }
 
-- 
1.5.6


  parent reply	other threads:[~2008-12-22 22:06 UTC|newest]

Thread overview: 66+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-12-22 21:47 [git patches] Ocfs2 patches for merge window, batch 2/3 Mark Fasheh
2008-12-22 21:47 ` [PATCH 01/56] quota: Add callbacks for allocating and destroying dquot structures Mark Fasheh
2008-12-22 21:47 ` [PATCH 02/56] quota: Increase size of variables for limits and inode usage Mark Fasheh
2008-12-22 21:47 ` [PATCH 03/56] quota: Remove bogus 'optimization' in check_idq() and check_bdq() Mark Fasheh
2008-12-22 21:47 ` [PATCH 04/56] quota: Make _SUSPENDED just a flag Mark Fasheh
2008-12-22 21:47 ` [PATCH 05/56] quota: Allow to separately enable quota accounting and enforcing limits Mark Fasheh
2008-12-22 21:47 ` [PATCH 06/56] ext3: Use sb_any_quota_loaded() instead of sb_any_quota_enabled() Mark Fasheh
2008-12-22 21:47 ` [PATCH 07/56] ext4: " Mark Fasheh
2008-12-22 21:47 ` [PATCH 08/56] reiserfs: " Mark Fasheh
2008-12-22 21:47 ` [PATCH 09/56] quota: Remove compatibility function sb_any_quota_enabled() Mark Fasheh
2008-12-22 21:47 ` [PATCH 10/56] quota: Introduce DQUOT_QUOTA_SYS_FILE flag Mark Fasheh
2008-12-22 21:47 ` [PATCH 11/56] quota: Move quotaio_v[12].h from include/linux/ to fs/ Mark Fasheh
2008-12-22 21:47 ` [PATCH 12/56] quota: Split off quota tree handling into a separate file Mark Fasheh
2008-12-22 21:47 ` [PATCH 13/56] quota: Convert union in mem_dqinfo to a pointer Mark Fasheh
2008-12-22 21:47 ` [PATCH 14/56] quota: Allow negative usage of space and inodes Mark Fasheh
2008-12-22 21:47 ` [PATCH 15/56] quota: Keep which entries were set by SETQUOTA quotactl Mark Fasheh
2008-12-22 21:47 ` [PATCH 16/56] quota: Update version number Mark Fasheh
2008-12-22 21:47 ` [PATCH 17/56] quota: Add helpers to allow ocfs2 specific quota initialization, freeing and recovery Mark Fasheh
2008-12-22 21:47 ` [PATCH 18/56] quota: Implement function for scanning active dquots Mark Fasheh
2008-12-22 21:48 ` [PATCH 19/56] mm: Export pdflush_operation() Mark Fasheh
2008-12-23  0:01   ` Andrew Morton
2008-12-25  1:05     ` Mark Fasheh
2008-12-31 19:28       ` Mark Fasheh
2008-12-31 22:17         ` Joel Becker
2008-12-31 23:09           ` [Ocfs2-devel] " Mark Fasheh
2009-01-05 13:27             ` Jan Kara
2008-12-22 21:48 ` [PATCH 20/56] ocfs2: Support nested transactions Mark Fasheh
2008-12-22 21:48 ` [PATCH 21/56] ocfs2: Assign feature bits and system inodes to quota feature and quota files Mark Fasheh
2008-12-22 21:48 ` [PATCH 22/56] ocfs2: Mark system files as not subject to quota accounting Mark Fasheh
2008-12-22 21:48 ` [PATCH 23/56] ocfs2: Implementation of local and global quota file handling Mark Fasheh
2008-12-23  0:11   ` Andrew Morton
2008-12-25  0:29     ` Mark Fasheh
2008-12-26  2:29       ` Andrew Morton
2008-12-22 21:48 ` [PATCH 24/56] ocfs2: Add quota calls for allocation and freeing of inodes and space Mark Fasheh
2008-12-22 21:48 ` [PATCH 25/56] ocfs2: Implement quota syncing thread Mark Fasheh
2008-12-22 21:48 ` [PATCH 26/56] ocfs2: Implement quota recovery Mark Fasheh
2008-12-22 21:48 ` [PATCH 27/56] ocfs2: Enable quota accounting on mount, disable on umount Mark Fasheh
2008-12-22 21:48 ` [PATCH 28/56] jbd2: Add BH_JBDPrivateStart Mark Fasheh
2008-12-22 21:48 ` [PATCH 29/56] ocfs2: Use BH_JBDPrivateStart instead of BH_Unshadow Mark Fasheh
2008-12-22 21:48 ` [PATCH 30/56] ocfs2: Add missing initialization Mark Fasheh
2008-12-22 21:48 ` [PATCH 31/56] ocfs2: Fix ocfs2_read_quota_block() error handling Mark Fasheh
2008-12-22 21:48 ` [PATCH 32/56] ocfs2: Fix oops when extending quota files Mark Fasheh
2008-12-22 21:48 ` [PATCH 33/56] ocfs2: Make ocfs2_get_quota_block() consistent with ocfs2_read_quota_block() Mark Fasheh
2008-12-22 21:48 ` [PATCH 34/56] ocfs2: Fix build warnings (64-bit types vs long long) Mark Fasheh
2008-12-22 21:48 ` [PATCH 35/56] quota: Unexport dqblk_v1.h and dqblk_v2.h Mark Fasheh
2008-12-22 21:48 ` [PATCH 36/56] quota: Export dquot_alloc() and dquot_destroy() functions Mark Fasheh
2008-12-22 21:48 ` [PATCH 37/56] reiserfs: Add default allocation routines for quota structures Mark Fasheh
2008-12-22 21:48 ` [PATCH 38/56] ext3: " Mark Fasheh
2008-12-22 21:48 ` [PATCH 39/56] ext4: " Mark Fasheh
2008-12-22 21:48 ` [PATCH 40/56] ocfs2: fix indendation in ocfs2_dquot_drop_slow Mark Fasheh
2008-12-22 21:48 ` [PATCH 41/56] ocfs2/quota: sparse fixes for quota Mark Fasheh
2008-12-22 21:48 ` [PATCH 42/56] ocfs2: Dirty the entire bucket in ocfs2_bucket_value_truncate() Mark Fasheh
2008-12-22 21:48 ` [PATCH 43/56] ocfs2: Narrow the transaction for deleting xattrs from a bucket Mark Fasheh
2008-12-22 21:48 ` [PATCH 44/56] ocfs2: Dirty the entire first bucket in ocfs2_extend_xattr_bucket() Mark Fasheh
2008-12-22 21:48 ` [PATCH 45/56] ocfs2: Dirty the entire first bucket in ocfs2_cp_xattr_cluster() Mark Fasheh
2008-12-22 21:48 ` [PATCH 46/56] ocfs2: Explain t_is_new " Mark Fasheh
2008-12-22 21:48 ` Mark Fasheh [this message]
2008-12-22 21:48 ` [PATCH 48/56] ocfs2: Rename ocfs2_cp_xattr_cluster() to ocfs2_mv_xattr_buckets() Mark Fasheh
2008-12-22 21:48 ` [PATCH 49/56] ocfs2: ocfs2_mv_xattr_buckets() can handle a partial cluster now Mark Fasheh
2008-12-22 21:48 ` [PATCH 50/56] ocfs2: Use ocfs2_mv_xattr_buckets() in ocfs2_mv_xattr_bucket_cross_cluster() Mark Fasheh
2008-12-22 21:48 ` [PATCH 51/56] ocfs2: Start using buckets in ocfs2_adjust_xattr_cross_cluster() Mark Fasheh
2008-12-22 21:48 ` [PATCH 52/56] ocfs2: Pass buckets into ocfs2_mv_xattr_bucket_cross_cluster() Mark Fasheh
2008-12-22 21:48 ` [PATCH 53/56] ocfs2: Move buckets up into ocfs2_add_new_xattr_cluster() Mark Fasheh
2008-12-22 21:48 ` [PATCH 54/56] ocfs2: Move buckets up into ocfs2_add_new_xattr_bucket() Mark Fasheh
2008-12-22 21:48 ` [PATCH 55/56] ocfs2: Pass xs->bucket " Mark Fasheh
2008-12-22 21:48 ` [PATCH 56/56] ocfs2/quota: Add QUOTA in mlog_attribute Mark Fasheh

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1229982517-3455-48-git-send-email-mfasheh@suse.com \
    --to=mfasheh@suse.com \
    --cc=joel.becker@oracle.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=ocfs2-devel@oss.oracle.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®